Light pollution can only be managed where it can be measured, and measured on one common scale. The Sky Integrating Sphere calibrates light-pollution sensors against the natural night sky, many at a time, in the field. It turned scattered instruments into networks that now watch over protected areas from the Azores to Extremadura.
Artificial light is now detectable over 26.5% of the land surface between 59° N and 55° S. For nocturnal wildlife it overrides the cues of moon and stars. At least 62 seabird species are known to be affected by land-based lights, most visibly the fledglings that leave their island burrows for the first time at night, are drawn towards streetlights and fall to the ground.
Cory’s shearwater breeds almost entirely on the islands of Macaronesia, and the largest part of its population nests in the Azores. Each autumn the birds cross the equator to winter in the South Atlantic, off South America and southern Africa; some return by a western route that reaches the waters off Newfoundland before heading back to their colony.
The journey spans an ocean, but its most vulnerable moment is local: the fledgling’s first flight from the burrow, at night, when a single lamp can bring it down. Every measure against that, from shielding and dimming to seasonal blackouts, depends on knowing how much light there is, of what colour, where and when, on the islands where the birds breed.
Routes drawn after the tracks of Cory’s shearwaters from Selvagem Grande in Dell’Ariccia et al. (2018), Scientific Reports 8, 3376; simplified and not exact. Coastlines: Natural Earth.
Satellites observe the light leaving the Earth, roughly once a night, and measure poorly exactly where it matters: in dark places, and in the blue light of modern LEDs. A sensor on the ground can read the sky every few minutes.
But each sensor reports brightness through its zero point, the constant that ties its signal to the standard scale. Until now, every zero point was set in one of very few laboratories, one instrument at a time, under an artificial lamp unlike the sky. Its uncertainty, 0.044 mag/arcsec², absorbed a large part of the natural variation the networks were meant to detect.
Schematic week of zenith sky brightness. A cloudy night amplifies skyglow; a new lamp appears as a lasting step. One sample a night cannot tell the two apart, or say when either began.
That is the whole principle. Place an instrument beside a reference whose zero point is known, record simultaneous readings, and the difference between them is the difference between their zero points. The sky itself becomes the integrating sphere.
m = ZP − 2.5 log10(f − fD)
Sky brightness m in magnitudes per square arcsecond, from the detector’s signal f and dark frequency fD. ZP, the zero point, is the unknown that SIS recovers.
The difficulty lies in the word same. The night sky is not a lamp: twilight, the Moon, cloud, the Milky Way and the zodiacal light all change it, and two sensors see an identical sky only under conditions that are defined in numbers.
Those conditions come from a characterisation of the natural night sky built on 11 million measurements from 44 photometers at dark sites around the world. Each contaminant brightens the sky in its own way and is removed by a criterion drawn from astronomical geometry, not tuned to the data, so the same filters apply at any site and to any batch of instruments.
Remove them one at a time.
Illustrative reconstruction after Alarcon et al. (2021), The Astronomical Journal 162, 25: eleven months of zenith measurements at Roque de los Muchachos Observatory, La Palma. What survives every filter is the natural sky, measured there at 21.94 ± 0.04 mag/arcsec², the darkest of the 44 sites.
Filtered pairs accumulate night after night, stray readings from aircraft or headlights are rejected by consensus fitting (RANSAC), and every instrument’s zero point settles at once. In the original validation the uncertainty stopped improving after about 360 qualifying measurements, which a week at Teide Observatory provides in any season, at below 0.02 mag/arcsec²: less than half the laboratory figure.
Simulation of the convergence reported in Alarcon et al. (2021), Sensors 21, 5590. Top left: the laboratory-calibrated reference. The laboratory figure is the published uncertainty for TESS-W.
Once calibration no longer depends on a laboratory, a network is limited only by how many instruments can be installed. More than 600 have been calibrated with SIS in five years. SIS made a new class of instrument worth building: solar-powered, wireless sensors installed once, where there is no power or cabling, and left to measure for years. Two of them, SG-WAS and NeXT, are manufactured commercially.
Every reading is published openly and in real time, on one calibrated scale, at data.eelabs.eu. The portal runs the calibration itself, so a sensor in the field can be checked against a reference without being removed.
Each point is an instrument; positions are indicative. Rings mark the calibration site at Teide Observatory and the natural-sky reference at Roque de los Muchachos.
Corvo is the smallest island of the Azores: one village of about five hundred people, and cliffs where shearwaters nest. Since July 2021 ten photometers have measured its sky, one of them on the rim of the Caldeirão crater. The satellite cannot help here: its night-light product returns no values for Corvo, whose pixels are dominated by the sea.
One sensor gives one number. Ten give a map of where the village’s light goes, but only if they read on one scale. An error of a tenth of a magnitude in a single instrument paints a glow where there is none.
Ten sensors on one scale: the glow of the village fades towards the dark interior of the island.
Illustrative model of zenith sky brightness over Corvo, interpolated from ten sensors. Positions and values are indicative; the uncalibrated case uses zero-point errors of up to 0.1 mag/arcsec². Lines join points of equal brightness.
The natural sky is never still. Airglow, the faint light of the upper atmosphere, drifts across it in slow waves. Photometers less than about 30 km apart see the same waves at the same moment and agree to 0.01 mag/arcsec²; beyond 150 to 200 km they no longer do. A cloud would break that agreement within 3 km, so the parallax places the source at least 100 km up, above the mesosphere.
For a network this is useful. Every sensor on Corvo shares the same airglow, so what they have in common is the natural sky, and what remains at each site is light from the ground. Remove the shared sky and the village’s lighting stands alone, down to the moment it is dimmed.
Schematic night over Corvo; the diagram is not to scale. The shared ripple is airglow. The step is the village’s lighting dimmed by 75%, the reduction its streetlights allow in the critical weeks for seabirds. Coherence of nearby photometers as measured in Alarcon et al. (2021).
At Puerto Villareal, Extremadura, a network photometer recorded an abrupt brightening at the end of 2019, when a new streetlight was installed a few metres away. Continuous calibrated monitoring turns a change in the night sky into something that can be dated and attributed.
21.94mag/arcsec²
The natural night sky brightness, measured at Roque de los Muchachos Observatory, La Palma, the darkest of the 44 sites analysed. Restoring natural darkness becomes a target that can be audited and written into a management plan.
NeXT measures colour as well as brightness, and colour governs what light does to wildlife. The natural sky at Teide Observatory sits at a V–I colour index of 1.78; sodium and white LED lighting fall between 0.8 and 1.2.
Public authorities have installed photometers inside Timanfaya National Park, Lanzarote, to quantify the artificial light that reaches the park and assess its effects on fauna. Across Macaronesia, the networks were built to map darkness in Natura 2000 sites and around seabird colonies.
It removes the barrier that kept artificial light from being measured, compared and verified where it matters most for biodiversity. This is what exists now because of it.
Field calibration against the sky, then autonomous instruments in remote protected areas, then dense networks on one scale, then open public data at data.eelabs.eu, then evidence for conservation and lighting policy.